Accurate solar resource forecasting is important because of the inherent variability associated with solar energy and its significant impact on the cost for energy producers. The traditional method applied in solar irradiance forecasting involves two main phases, related to (1) data selection and (2) model selection, training, and evaluation. In this study, we propose a novel end-to-end methodology for solar irradiance forecasting that starts with the search for the data and all of the preprocessing operations involved in obtaining a quality dataset, continuing by imputing missing data with the BERT (Bidirectional Encoder Representations from Transformers) model, and ending with obtaining and evaluating the predicted values. This novel methodology is based on three phases; namely, Phase_1, related to the acquisition and preparation of the data, Phase_2, related to the proposed imputation with a BERT model, and Phase_3, related to the training and prediction with new models based on deep learning. These phases of the proposed methodology can be applied in a disjointed manner, and were used on two public datasets accessible to the scientific community. Each of the proposed phases proved to be valuable for the workflow, and the application of the novel method delivered increases in performance of up to 3 percentage points (3%) when compared to the traditional approach.

Studies addressing the supervised extraction of geospatial elements from aerial imagery with semantic segmentation operations (including road surface areas) commonly feature tile sizes varying from 256 × 256 pixels to 1024 × 1024 pixels with no overlap. Relevant geo-computing works in the field often comment on prediction errors that could be attributed to the effect of tile size (number of pixels or the amount of information in the processed image) or to the overlap levels between adjacent image tiles (caused by the absence of continuity information near the borders). This study provides further insights into the impact of tile overlaps and tile sizes on the performance of deep learning (DL) models trained for road extraction. In this work, three semantic segmentation architectures were trained on data from the SROADEX dataset (orthoimages and their binary road masks) that contains approximately 700 million pixels of the positive “Road” class for the road surface area extraction task. First, a statistical analysis is conducted on the performance metrics achieved on unseen testing data featuring around 18 million pixels of the positive class. The goal of this analysis was to study the difference in mean performance and the main and interaction effects of the fixed factors on the dependent variables. The statistical tests proved that the impact on performance was significant for the main effects and for the two-way interaction between tile size and tile overlap and between tile size and DL architecture, at a level of significance of 0.05. We provide further insights and trends in the predictions of the extensive qualitative analysis carried out with the predictions of the best models at each tile size. The results indicate that training the DL models on larger tile sizes with a small percentage of overlap delivers better road representations and that testing different combinations of model and tile sizes can help achieve a better extraction performance.

Popular geo-computer vision works make use of aerial imagery, with sizes ranging from 64 × 64 to 1024 × 1024 pixels without any overlap, although the learning process of deep learning models can be affected by the reduced semantic context or the lack of information near the image boundaries. In this work, the impact of three tile sizes (256 × 256, 512 × 512, and 1024 × 1024 pixels) and two overlap levels (no overlap and 12.5% overlap) on the performance of road classification models was statistically evaluated. For this, two convolutional neural networks used in various tasks of geospatial object extraction were trained (using the same hyperparameters) on a large dataset (containing aerial image data covering 8650 km2 of the Spanish territory that was labelled with binary road information) under twelve different scenarios, with each scenario featuring a different combination of tile size and overlap. To assess their generalisation capacity, the performance of all resulting models was evaluated on data from novel areas covering approximately 825 km2. The performance metrics obtained were analysed using appropriate descriptive and inferential statistical techniques to evaluate the impact of distinct levels of the fixed factors (tile size, tile overlap, and neural network architecture) on them. Statistical tests were applied to study the main and interaction effects of the fixed factors on the performance. A significance level of 0.05 was applied to all the null hypothesis tests. The results were highly significant for the main effects (p-values lower than 0.001), while the two-way and three-way interaction effects among them had different levels of significance. The results indicate that the training of road classification models on images with a higher tile size (more semantic context) and a higher amount of tile overlap (additional border context and continuity) significantly impacts their performance. The best model was trained on a dataset featuring tiles with a size of 1024 × 1024 pixels and a 12.5% overlap, and achieved a loss value of 0.0984, an F1 score of 0.8728, and an ROC-AUC score of 0.9766, together with an error rate of 3.5% on the test set.

In solar forecasting, it is common practice for solar data (be it irradiance or photovoltaic power) to be converted into a stationary index (e.g., clear-sky or clearness index) before being used as inputs for solar-forecasting models. However, its actual impact is rarely quantified. Thus, this paper aims to study the impact of including this processing step in the modeling workflow within the scope of very-short-term spatio-temporal forecasting. Several forecasting models are considered, and the observed impact is shown to be model-dependent. Persistence does not benefit from this for such short timescales; however, the statistical models achieve an additional 0.5 to 2.5 percentual points (PPs) in terms of the forecasting skill. Machine-learning (ML) models achieve 0.9 to 1.9 more PPs compared to a linear regression, indicating that stationarization reveals non-linear patterns in the data. The exception is Random Forest, which underperforms in comparison with the other models. Lastly, the inclusion of solar elevation and azimuth angles as inputs is tested since these are easy to compute and can inform the model on time-dependent patterns. Only the cases where the input is not made stationary, or the underperforming Random Forest model, seem to benefit from this. This indicates that the apparent Sun position data can compensate for the lack of stationarization in the solar inputs and can help the models to differentiate the daily and seasonal variability from the shorter-term, weather-driven variability.

Representative road lines (e.g., continuous and discontinuous lane markings, or lines defining highway entrances or exits) found on highway pavement represent valuable information for the generation of high-definition road cartography. A correct extraction of these road markings is becoming an increasingly important aspect in the context of accurate road decision support systems creation (used by autonomous driving solutions). In this work, the extraction of the mentioned representative road lines is tackled as a multiclass, supervised semantic segmentation task with deep learning methods. In this regard, thirteen semantic segmentation models (based on the state-of-the-art LinkNet, PSPNet (Pyramid Scene Parsing Network), FPN (Feature Pyramid Network), U-Net and U-Net++, and MA-Net (Multi-Attention-Network) architectures) were trained on a novel dataset (named "WhiteRoadLines") containing 27,025 images with a spatial resolution of 15 cm, labelled at pixel level. Post-training, the performance was evaluated using appropriate performance metrics; the best performing model being a version of U-Net coupled with Inception-ResNet-v2 that achieved a mean Intersection-over-Union score of 0.6559, while delivering high quality predictions on unseen, test data. The results proved the appropriateness of using DL techniques for completing high-definition road cartography with information related to road markings found on highways.

In this work, three new convolutional neural network models—spatio-temporal convolutional neural network versions 1 and 2 (ST_CNN_v1 and ST_CNN_v2), and the spatio-temporal dilated convolutional neural network (ST_Dilated_CNN)—are proposed for solar forecasting and processing global horizontal irradiance (GHI) data enriched with meteorological and astronomical variables. A comparative analysis of the proposed models with two traditional benchmark models shows that the proposed ST_Dilated_CNN model outperforms the rest in capturing long-range dependencies, achieving a mean absolute error of 31.12 W/m2, a mean squared error of 54.07 W/m2, and a forecast skill of 37.21%. The statistical analysis carried out on the test set suggested highly significant differences in performance (p-values lower than 0.001 for all metrics in all the considered scenarios), with the model with the lowest variability in performance being ST_CNN_v2. The statistical tests applied confirmed the robustness and reliability of the proposed models under different conditions. In addition, this work highlights the significant influence of astronomical variables on prediction performance. The study also highlights the intricate relationship between the proposed models and meteorological and astronomical input characteristics, providing important insights into the field of solar prediction and reaffirming the need for further research into variability factors that affect the performance of models.

Active learning methodologies are oducational strategies where students take a proactive role, and where they must meaningfully work and d reflect on the competencies of the academic units. More than twelve different active methodologics have been developod in recent years. Selecting one or another usually depends on the available time and physical space, the nature of the academic unit to be learned, or the number of students. But many authors also highlight the comfort level of professors as a key factor to understand the adoption of some active strategies over others. This article aims to study how this "comfort level" affects the adoption of the different active learning methodologies, in engineering education conters, and using the Rogers Diffusion of Innovation theory. A collection of survey questions was conducted among engineering professors at Universidad Polit & eacute;cnica de Madrid and Universidad Alfonso X el Sabio. Responses were based on the Likert scale and statistical methods were used to analyze the collected data. Results confirmod that the relative advantage has the highest positive impact on the adoption rate and observability. However, complexity, the nature of the social system and the type of innovation-decision have a negative impact on the adoption rate. Engineering education centers ARE suggested to increase the social recognition of active learning adoption to improve the adoption rate.

The availability of solar irradiance time series without missing data is an ideal scenario for researchers in the field. However, it is not achievable for a variety of reasons, such as measurement errors, sampling gaps, or other factors. Time series imputation methods can be a solution to the lack of data and, in this paper, we study the applicability of Bidirectional Encoder Representations from Transformers (BERT) as an irradiance time series imputation solution. In this regard, a BERT model was trained from scratch for the masked language modelling (MLM) task, and the quality of the imputation was evaluated according to the number of missing values and the position within the series. The experiments were conducted over a dataset of 165 stations, captured by meteorological stations distributed over the Spanish regions of Galicia, Castile, and León. In the evaluation process, an average coefficient of determination (R2 score) of 0.89% was obtained, the maximum result being 0.95%.

The dataset consists of 27,025 PNG images (256x256 pixels) of high resolution aerial orthoimages at 0,15 m/pixel of resolution. The images contain information related to representative road lines and markings found on highway pavement and is labelled for multi-class semantic segmentation with tree classes of white roadlines and markings: (1) continuous line (black color), (2) dashed line (dark gray color) and (3) separation of entry and exit lanes (light gray color), together with (4) the background (white color). The dataset has been created in the framework of the SROADEX project to train a multiclass semantic segmentation process based on Deep Learning.The data have been generated using scripts developed in Python using Open Source libraries (GDAL/OGR and MapScript) for rasterization of vector cartography representing the three different types of white road lines. This cartography has been obtained from Spanish official sources (National Geographic Institute) that we haverevised and edited in a meticulous and systematic way to verify that the road lines are represented on the cartography according to the orthoimages, available on January 1, 2022 in the download center of the National Center of Geographic Information (CNIG). In the digitisation process, 46 homogeneously distributed areas of Spain have been selected. The orthoimages used have been resampled from the original resolution of 0,25m/pixel to 0,15m/pixel, as this is closer to the width of two of the three classes of white lines in the dataset. It resulted in 80% of the images for training (21622), 10% for validation (2702) and 10% for testing (2701). The following table summarises the number of pixels of each category included in each of the three sub-datasets Set Nº images Class_1 (continuous line) Class_2 (discontinuous line) Class_3 (line defining highway entrance or exit) Class_4 (background) Train 21,622 27,633,537 4,543,552 3,284,380 1,381,557,923 Validation 2,702 3,433,103 570,741 395,646 172,678,782 Test 2,701 3,435,072 536,838 429,527 172,611,299 Total 27,025 34,501,712 5,651,131 4,109,553 1,726,848,004

Along with the development and improvement of measuring technologies and techniques in recent times, new methods have appeared to model and monitor the behavior of land and constructions over time. The main purpose of this research was to develop a new methodology to model and monitor large buildings in a non-invasive way. The methods proposed in this research are non-destructive and can be used to monitor the behavior of buildings over time. A method of comparing point clouds obtained using terrestrial laser scanning combined with aerial photogrammetric methods was used in this study. The advantages and disadvantages of using non-destructive measurement techniques over the classic methods were also analyzed. With a building located in the University of Agricultural Sciences and Veterinary Medicine Cluj-Napoca campus as a case study and with the help of the proposed methods, the deformations over time of the facades of that building were determined. As one of the main conclusions of this case study, it can be stated that the proposed methods are adequate to model and monitor the behavior of constructions over time, ensuring a satisfactory degree of precision and accuracy. The methodology can be successfully applied to other similar projects.

Accurate solar forecasting lately relies on advances in the field of artificial intelligence and on the availability of databases with large amounts of information on meteorological variables. In this paper, we present the methodology applied to introduce a large-scale, public, and solar irradiance dataset, CyL-GHI, containing refined data from 37 stations found within the Spanish region of Castile and León (Spanish: Castilla y León, or CyL). In addition to the data cleaning steps, the procedure also features steps that enable the addition of meteorological and geographical variables that complement the value of the initial data. The proposed dataset, resulting from applying the processing methodology, is delivered both in raw format and with the quality processing applied, and continuously covers 18 years (the period from 1 January 2002 to 31 December 2019), with a temporal resolution of 30 min. CyL-GHI can result in great importance in studies focused on the spatial-temporal characteristics of solar irradiance data, due to the geographical information considered that enables a regional analysis of the phenomena (the 37 stations cover a land area larger than 94,226 km2). Afterwards, three popular artificial intelligence algorithms were optimised and tested on CyL-GHI, their performance values being offered as baselines to compare other forecasting implementations. Furthermore, the ERA5 values corresponding to the studied area were analysed and compared with performance values delivered by the trained models. The inclusion of previous observations of neighbours as input to an optimised Random Forest model (applying a spatio-temporal approach) improved the predictive capability of the machine learning models by almost 3%.

Most existing road extraction approaches apply learning models based on semantic segmentation networks and consider reduced study areas, featuring favorable scenarios. In this work, an end-to-end processing strategy to extract the road surface areas from aerial orthoimages at the scale of the national territory is proposed. The road mapping solution is based on the consecutive execution of deep learning (DL) models trained for ① road recognition, ② semantic segmentation of road surface areas, and ③ post-processing of the initial predictions with conditional generative learning, within the same processing environment. The workflow also involves steps such as checking if the aerial image is found within the country’s borders, performing the three mentioned DL operations, applying a p=0.5 decision limit to the class predictions, or considering only the central 75% of the image to reduce prediction errors near the image boundaries. Applying the proposed road mapping solution translates to operations aimed at checking if the latest existing cartographic support (aerial orthophotos divided into tiles of 256 × 256 pixels) contains the continuous geospatial element, to obtain a linear approximation of its geometry using supervised learning, and to improve the initial semantic segmentation results with post-processing based on image-to-image translation. The proposed approach was implemented and tested on the openly available benchmarking SROADEX dataset (containing more than 527,000 tiles covering approximately 8650 km2 of the Spanish territory) and delivered a maximum increase in performance metrics of 10.6% on unseen, testing data. The predictions on new areas displayed clearly higher quality when compared to existing state-of-the-art implementations trained for the same task.

Having a complete and high-quality geospatial catalogue of existing large-scale photovoltaic (PV) panels is very important nowadays, due to the rapid increase in the use of this type of installations. This catalogue could be used to estimate, with a higher level of granularity, the energy that can be produced from solar radiation forecasts, to operate the electricity system efficiently by adjusting supply to demand, as well as for fiscal reasons. However, most existing works in the specialized literature focus on the extraction of PV panels in reduced, favourable scenes. In this study, a processing strategy to obtain PV panel arrays geometries from aerial orthoimages at a very large scale, is proposed. The processing strategy includes operations for PV panel array classification and semantic extraction, and algorithmic improvement and simplification of the vectorization results. The processing workflow was trained and tested on two datasets containing aerial images of 256 × 256 pixels with the studied polygonal feature at two different spatial resolutions (with more than 185,000 and 690,000 images for spatial resolutions of 0.5 m and 0.25 m, respectively). The image tiles were labelled at pixel level with PV panel information for both spatial resolutions and cover representative, extended urban and rural regions of Spain. The accuracies achieved in the PV panel arrays classification are superior to 0.999, while the semantic segmentation performance is superior to 0.90 in the Intersection over Union score. The assessment of the proposed procedure of fifteen new, unseen areas distributed over the Spanish territory (covering approximately 8125 km2) demonstrated its suitability (in the analysis, percentage of incompletely detected installations of only 11 % were achieved) and indicated that future works focused on the improvement of the initial predictions are should be carried out.

The connection between solar irradiance and wind is a topic of interest in the field of renewable energies, as wind data have proven to be effective predictors of solar energy, being indicators of cloud movement and atmospheric conditions. This study focuses the use of decision tree-based algorithms (random forest, XGBoost and light gradient boosting machine, and LightGBM) to analyse the impact of the meridional and zonal wind components as input variables. In the study, past observations of neighbours were included as predictors to include a spatiotemporal analysis. The studied models were trained on the open, well-established OIH dataset (containing data from Oahu Island, Hawaii, located in the United States of America) featuring predominantly northeasterly winds. In the post-training analysis, it was found that the inclusion of the wind components resulted in a mean improvement of approximately 1% in the forecast skill (FS) score for all models, with the XGBoost model being the best performing model (with a 27.63% FS score).

The data have been generated using scripts developed in Python with Open-Source libraries (GDAL/OGR and MapScript) to rasterize of vector cartography representing the photovoltaic (PV) panels instalations in urban, industrial, and rural areas. This PV panels cartography has been generated by manual digitalizing the PV panels found latest aerial orthofotographs available on June 1, 2021 from Plano Nacional de Ortofotografía Aérea (PNOA), produced by the National Geographic Institute of Spain, using the Web Map Service PNOA-MA. The dataset consists of 239,680 images of 256 × 256 pixels in size, in png format, labelled with Class_1: “Contains PV panel” and Class_2: “Does not contain PV panel”, that were pre-divided with a split criterion of 70:10:20%. in train, validation and test folders, respectively. The structure of the data is as follows: 1-Panels-Ortho and 1-Panels-Mask contain the images featuring PV panels and their corresponding ground truth mask for training the semantic segmentation networks. 1-Panels-Ortho and 2-NoPanels-Ortho contain images containing and not containing PV panels, for the training of binary recognition models of PV panels. Moreover, in each folder the structure is the same: train, test, validation containing 70%, 10% and 20% of the total images and masks of each type. 1-Panels-Ortho |----Train |----Test -----Validation 1-Panels-Mask |----Train |----Test -----Validation 2-NoPanels-Ortho |----Train |----Test -----Validation

Arrow signs found on roadway pavement are an important component of modern transportation systems. Given the rise in autonomous vehicles, public agencies are increasingly interested in accurately identifying and analysing detailed road pavement information to generate comprehensive road maps and decision support systems that can optimise traffic flow, enhance road safety, and provide complete official road cartographic support (that can be used in autonomous driving tasks). As arrow signs are a fundamental component of traffic guidance, this paper aims to present a novel deep learning-based approach to identify the orientation and direction of arrow signs on marked roadway pavements using high-resolution aerial orthoimages. The approach is based on convolutional neural network architectures (VGGNet, ResNet, Xception, and DenseNet) that are modified and adapted for regression tasks with a proposed learning structure, together with an ad hoc model, specially introduced for this task. Although the best-performing artificial neural network was based on VGGNet (VGG-19 variant), it only slightly surpassed the proposed ad hoc model in the average values of the R2 score, mean squared error, and angular error by 0.005, 0.001, and 0.036, respectively, using the training set (the ad hoc model delivered an average R2 score, mean squared error, and angular error of 0.9874, 0.001, and 2.516, respectively). Furthermore, the ad hoc model’s predictions using the test set were the most consistent (a standard deviation of the R2 score of 0.033 compared with the score of 0.042 achieved using VGG19), while being almost eight times more computationally efficient when compared with the VGG19 model (2,673,729 parameters vs VGG19′s 20,321,985 parameters).

Accurate information on surface soil moisture (SSM) content at a global scale under different climatic conditions is important for hydrological and climatological applications. Machine-learning-based systematic integration of in situ hydrological measurements, complex environmental and climate data, and satellite observation facilitate the generation of reliable data products to monitor and analyse the exchange of water, energy, and carbon in the Earth system at a proper space–time resolution. This study investigates the estimation of daily SSM using 8 optimised machine learning (ML) algorithms and 10 ensemble models (constructed via model bootstrap aggregating techniques and five-fold cross-validation). The algorithmic implementations were trained and tested using International Soil Moisture Network (ISMN) data collected from 1722 stations distributed across the world. The result showed that the K-neighbours Regressor (KNR) had the lowest root-mean-square error (0.0379 cm3 cm−3) on the “test_random” set (for testing the performance of randomly split data during training), the Random Forest Regressor (RFR) had the lowest RMSE (0.0599 cm3 cm−3) on the “test_temporal” set (for testing the performance on the period that was not used in training), and AdaBoost (AB) had the lowest RMSE (0.0786 cm3 cm−3) on the “test_independent-stations” set (for testing the performance on the stations that were not used in training). Independent evaluation on novel stations across different climate zones was conducted. For the optimised ML algorithms, the median RMSE values were below 0.1 cm3 cm−3. GradientBoosting (GB), Multi-layer Perceptron Regressor (MLPR), Stochastic Gradient Descent Regressor (SGDR), and RFR achieved a median r score of 0.6 in 12, 11, 9, and 9 climate zones, respectively, out of 15 climate zones. The performance of ensemble models improved significantly, with the median RMSE value below 0.075 cm3 cm−3 for all climate zones. All voting regressors achieved r scores of above 0.6 in 13 climate zones; BSh (hot semi-arid climate) and BWh (hot desert climate) were the exceptions because of the sparse distribution of training stations. The metric evaluation showed that ensemble models can improve the performance of single ML algorithms and achieve more stable results. Based on the results computed for three different test sets, the ensemble model with KNR, RFR and Extreme Gradient Boosting (XB) performed the best. Overall, our investigation shows that ensemble machine learning algorithms have a greater capability with respect to predicting SSM compared with the optimised or base ML algorithms; this indicates their huge potential applicability in estimating water cycle budgets, managing irrigation, and predicting crop yields.

To better forecast solar variability, spatio-temporal methods exploit spatially distributed solar time series, seeking to improve forecasting accuracy by including neighboring solar information. This review work is, to the authors’ understanding, the first to offer a compendium of references published since 2011 on such approaches for global horizontal irradiance and photovoltaic generation. The identified bibliography was categorized according to different parameters (method, data sources, baselines, performance metrics, forecasting horizon), and associated statistics were explored. Lastly, general findings are outlined, and suggestions for future research are provided based on the identification of less explored methods and data sources.

In this data paper, we propose an open dataset (named SROADEX) containing more than 527,000 image tiles of 256 × 256 pixels stored in the lossless PNG format, tagged at pixel level with road information. The dataset covers approximately 8650 km2 of the Spanish territory, is divided in train, validation and test sets and can be used by researchers and professionals for training other extraction solutions and benchmarking additional models. The SROADEX dataset is available under a CC-BY 4.0 licence and can be freely downloaded from the Zenodo repository.

In addition to machine learning methods, deep learning is increasingly being used for solar radiation prediction. In this study, two new deep learning models with attention mechanisms are proposed, and the influence of the temporal representation on the short-term results is evaluated. To address this question, the two networks were trained on an irradiation dataset from a network of 80 meteorological stations in Galicia, Spain, with different representation intervals being simulated. The results show that, although the proposed models achieve better results in comparison to other popular models trained for the same task, the transformation performed on the data degrades the prediction ability of the models as the representation interval increases.